mwangiernest552-cloud/urban-flood-health-kenya

# 🌊 Urban Flooding & Health: AI-Powered Risk Prediction for Kenya > An AI-driven early warning system that predicts flood-induced health risks in Nairobi's urban settlements 24-72 hours in advance, enabling targeted public health interventions. --- ## 🎯 Problem Statement Kenya's urban centers face intensifying flood-health crises. The **March-April-May 2024 rains** caused Kenya's worst flooding in decades: - 🏠 **95,859 households** affected - 👨‍👩‍👧‍👦 **52,673 families** displaced - 💔 **284 fatalities** - 🏥 **42 health facilities** overwhelmed - 🚰 **162 water sources** destroyed - 🏫 **285 schools** damaged **Key Statistics:** - 27% of Nairobi's total area is flood-prone - 35% of built-up areas exposed to flooding - 54% of formal areas at risk - Up to 40% of informal settlements (Korogocho, Kibera, Mathare) vulnerable Current early warning systems lack actionable granularity and health integration. This project bridges that gap. --- ## 🚀 Solution Overview A **Random Forest machine learning model** that: - ✅ Predicts flood-induced health risk 24-72 hours ahead - ✅ Identifies spatial hotspots at sub-county level (17 constituencies) - ✅ Achieves **70%+ accuracy** on 2021-2024 test data - ✅ Validated against 2024 MAM floods (worst on record) - ✅ Provides actionable risk levels for public health interventions --- ## 📊 Model Performance | Metric | Score | |--------|-------| | **Test Accuracy** | 70%+ | | **Precision** | High | | **Recall** | High | | **2024 MAM Validation** | Successfully detected worst flooding in decades | | **Features Used** | 9 predictors (rainfall + seasonal) | | **Training Period** | 1981-2020 (40 years of data) | | **Testing Period** | 2021-2024 (including 2024 floods) | **Top Predictive Features:** 1. 🌧️ 30-day cumulative rainfall 2. 💧 Antecedent Precipitation Index (soil moisture proxy) 3. 🌦️ 20-day cumulative rainfall 4. 📅 Seasonal indicators (MAM & OND rainy seasons) --- ## 🗂️ Project Structure